Artificial intelligence methodology for separation andclassification of partial discharge signals

摘要

Results of investigations performed in order to improve thecurrent diagnostic techniques used for the evaluation of insulationsystems of HV apparatus are presented in this paper. Improvements comefrom the development of a new measuring system which allows the digitalacquisition of Partial Discharge (PD) signals and a new separationmethod, based on a Fuzzy Classifier, for the analysis of the PD-pulseshape signals. The identification of the classes, relevant to differentPD phenomena, is then performed by means of PD-pulse height and phaseanalysis. The proposed approach is supported by the analysis of PD dataobtained from insulation systems of stator bars withartificially-reproduced defects